What’s in a Face? - Michigan State University

Ó 2013 Federation of European Psychophysiology Societies

Article

What’s in a Face?

The Late Positive Potential Reflects the Level of Facial
Affect Expression

Elizabeth R. Duval,1,5 Jason S. Moser,2 Jonathan D. Huppert,3 and
Robert F. Simons4

1Department of Psychiatry, University of Michigan, MI, USA, 2Department of Psychology, Michigan
State University, East Lansing, MI, USA, 3Department of Psychology, Hebrew University, Jerusalem,

Israel, 4Department of Psychology, University of Delaware, Newark, NJ, USA, 5VA Ann Arbor
Healthcare System, Ann Arbor, MI, USA

Abstract. Morphed faces depicting varying degrees of affect expression can be used to investigate the processing of ambiguous and thus more
ecologically valid facial stimuli. Event-related brain potentials (ERPs) were measured while participants viewed a series of faces ranging in 10%
increments from prototypically happy to prototypically neutral to prototypically angry. Results revealed that the late positive potential (LPP) – an
ERP reflecting later stages of stimulus processing – followed the degree of expression of facial affect such that faces depicting a greater amount of
affect elicited larger LPPs as compared to faces depicting less affect. The enhanced LPP to faces depicting greater amounts of affect was also
more sustained for angry compared to happy faces – in general, angry faces elicited larger LPPs. Overall, these results demonstrate the sensitivity
of the LPP to more ecologically valid facial expressions and thus the visual system’s finely tuned discriminability of important social
signals.

Keywords: event-related potentials, late positive potential, affective face processing, morphing

Faces convey a rich array of information about individuals responses to more ecologically valid emotional expressions
(age, gender, ethnicity, emotional state) and the environment is to morph prototypical expressions to create stimuli depict-
(approach/avoid intentions). Facial expressions of emotion ing varying degrees of facial affect. Recent research suggests
are particularly relevant to the survival of social animals. that healthy individuals seem quite accurate at identifying
The circumplex model of emotion (Posner, Russell, & the primary emotion displayed by morphed faces. Signal
Peterson, 2005; Russell, 1980) provides a parsimonious detection methods reveal that when presented with faces
structure for organizing facial expressions of emotion by depicting varying degrees of emotional expression along
positing that all emotions derive from two higher-order, angry-neutral, fear-neutral, and angry-fear spectrums, adults
orthogonal, dimensions: hedonic valence and arousal. were able to accurately identify facial emotion along a con-
Valence refers to the degree of pleasure a stimulus evokes tinuum (Thomas, De Bellis, Graham, & LaBar, 2007).
and arousal refers to the level of activation evoked. This
organization of emotions serves as the basis for much of Little is known, however, about the underlying neural
the research on facial expressions of emotion and thus rep- mechanisms involved in the possessing of morphed – that
resents the foundation for the present study. is, more ecologically valid – facial expressions. Event-
related brain potentials (ERPs) provide an ideal complement
Despite the omnipresence of faces and their survival to behavioral measures because their excellent temporal res-
value, many studies examining face processing use proto- olution allows for the examination of the sequence of con-
typical faces, depicting emotional extremes. However, facial stituent neural operations involved in stimulus- and
expressions in daily life are fluid and consist of a multitude response-processing of the order of milliseconds. In this
of combinations and degrees of affect expression. Therefore, way, ERPs aid in separating out stimulus- from response-
the facial expressions processed in every-day social situa- related neural processes that are confounded in behavioral
tions are often more ambiguous and dynamic than the pro- measures. A number of studies have already demonstrated
totypes used in the laboratory. One method to investigate ERP modulations during the processing of faces depicting

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38
DOI: 10.1027/0269-8803/a000083

28 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

prototypical emotional expressions (see Eimer & Holmes, reflecting different emotions compared to the same emotion
2007 for a review).1 For instance, several studies have re- (Campanella et al., 2002), and was smaller and slower in de-
ported an early-to-middle latency (< 300 ms) positive shift pressed participants compared to controls for all faces, espe-
in frontal regions in response to prototypical emotional (po- cially morphed faces displaying low levels of happy affect
sitive and negative) compared to neutral faces, indicative of (Cavanagh & Geisler). Sprengelmeyer and Jentzsch also
the recognition of emotional expression by areas of the fron- found that ERP positivities were larger between 200 and
tal cortex (Eimer & Holmes, 2007). Studies reporting on the 600 ms for morphed faces displaying greater amounts of af-
modulation of later (> 300 ms) ERP components also find fect compared to neutral faces, with faces depicting interme-
an effect of prototypical facial affect. The Late Positive Po- diate levels of affect resulting in intermediate increases in
tential (LPP) is of particular interest in the current study be- the ERP. However, these studies examined only three morph
cause it is sensitive to the emotional content of a broad range levels (Cavanagh & Geisler, 2006; Sprengelmeyer &
of stimuli (Cuthbert, Schupp, Bradley, Birbaumer, & Lang, Jentzsch, 2006) or presented two faces at the same time
2000). The LPP is a centro-parietally maximal positive shift within the context of a discrimination task (Campanella
that reaches its maximum amplitude between 300 and et al., 2002). Thus, the findings to date do not provide a clear
1,000 ms following stimulus onset. It has been proposed understanding of ERP modulations during, and therefore the
to index attention, perception, and memory processes elic- information processing stages involved in, the viewing of a
ited by various types of motivationally relevant stimuli continuum of morphed faces.
(Bradley, 2009; Donchin, 1981; Donchin & Coles, 1988;
Nieuwenhuis, Aston-Jones, & Cohen, 2005, Schupp et al., The purpose of the current study, then, was to build upon
2000). previous findings by investigating the modulation of early
and late positivities, including the LPP, during the viewing
To date, several studies (e.g., Eimer, Holmes, & of morphed affective faces. To extend previous studies, a
McGlone, 2003) suggest that the LPP is particularly sensi- broader range of facial expressions were created depicting
tive to the arousal quality of emotional faces, reporting that varying levels of affective expression. This resulted in
the LPP is equally enhanced to both prototypical positive prototypical angry, neutral, and happy faces with 18 addi-
and negative facial expressions compared to neutral faces. tional faces depicting 10% differences in affective expres-
Schupp et al. (2004), on the other hand, suggest that the sion along the happy-neutral and angry-neutral spectrum.
LPP may also be sensitive to the valence of stimuli. They Participants also rated stimuli using the Self-Assessment
demonstrated larger LPPs during the processing of prototyp- Manikin (SAM; Bradley & Lang, 1994) to assess both va-
ical angry faces compared to prototypical happy and lence and arousal dimensions on a 1 (very unpleasant;
prototypical neutral faces, consistent with the so-called low arousing) to 9 (very pleasant; highly arousing) scale.
‘‘negativity bias hypothesis.’’ This hypothesis states that We predicted an effect of arousal on ERP components. Spe-
threatening/negative faces are processed more thoroughly cifically, we predicted that the happiest and angriest expres-
than friendly/positive faces, which has been supported in a sions would result in the largest increase in ERP positivity
number of other studies using both real and schematic pro- measured at frontal recording sites in early time windows
totypical faces (Ito, Larsen, Smith, & Cacioppo, 1998; (< 300 ms; Eimer & Holmes, 2007) and at parietal record-
Kanouse & Hansen, 1971; Ohman, Lundqvist, & Esteves, ing sites during the LPP time windows (> 300 ms; Schupp
2001; Peeters & Czapinski, 1990; Skowronski & Carlston, et al., 2004) and the most neutral faces would result in the
1989; Tipples, Atkinson, & Young, 2002). smallest ERP positivity. Intermediate faces depicting vary-
ing degrees of happiness and anger would result in an in-
More important to the current investigation, all face pro- crease in ERP positivity consistent with the degree of
cessing studies reviewed above used faces portraying proto- affect expressed. Given past results supporting the negativity
typical affective expressions that depict the extreme end of bias hypothesis (e.g., Schupp et al., 2004), we also consid-
the emotion. Only a few studies have investigated whether ered the possibility of a valence effect in the LPP, such that
morphed faces modulate early and late ERP components a larger LPP would be elicited during the processing of an-
(Achaibou, Pourtois, Schwartz, & Vuilleumier, 2008; gry compared to happy and neutral faces in general.
Campanella, Quinet, Crommelinck, & Guerit, 2002;
Cavanagh & Geisler, 2006; Sprengelmeyer & Jentzsch, For SAM valence ratings, we predicted that the happiest
2006). The results of previous studies using morphed faces faces would be rated as the most positive, the angriest faces
suggest that both early and late ERP components are mod- would be rated as the most negative, and ratings for interme-
ulated by the degree of facial affect expression. Campanella diate faces would coincide with the level of affect expressed.
et al. (2002), Cavanagh and Geisler (2006), and Sprengel- For SAM arousal ratings, we predicted that faces depicting
meyer and Jentzsch (2006) reported ERP modulations greater amounts of affective expression (both positive and
between 200 and 600 ms while participants viewed faces negative) would be rated as the most arousing, with ratings
depicting varying degrees of affective expression. Specifi- for intermediate faces coinciding with the level of affect
cally, the P3 was larger when participants viewed two faces expressed.

1 The N170 is often analyzed within the context of face processing, but it is unclear at present if it indexes affective or structural aspects of
face stimuli (Eimer & Holmes, 2007). Because the primary purpose of the current study was to examine early and late ERP positivities that
show consistent modulations by emotional faces, our review of the existing literature focuses exclusively on these components.

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing

E. R. Duval et al.: LPP Reflects Differences in Facial Affect 29

Table 1. Average number of control points used for morphing Average
17.0
Facial region Happy/neutral morph Angry/neutral morph 14.1
22.3
Eyebrows 17.0 17.0 12.0
Eyes 14.0 14.2 11.0
Mouth 24.0 20.6 6.8
Nose 12.0 12.0 46.0
Cheek/chin 11.0 11.0 20.6
Forehead 6.8 6.8
Face circumference 46.0 46.0
Hair/neck 20.5 20.6

Method well, to smooth out all facial features. An average of
149.8 control points was used for each face. The average
Participants number of control points for the angry-neutral morphs
(148.3) was comparable to that used for the happy-neutral
Twenty-nine undergraduate students (16 female) recruited morphs (151.4; t(7) = .90, p > .05). For each of the original
through the University of Delaware Psychology Department images, the sets of control points resulted in the formation of
participant pool took part in the current study for course grids over each of the faces, which meshed during morphing
credit. to create the composite images (see Steyvers for a more de-
tailed description of mesh warping). The final product was a
Stimuli and Morphing Procedures set of 21 images for each model (210 total images; 105 male
and 105 female) varying on a continuum from happy to neu-
In line with the circumplex model of emotion (Posner et al., tral to angry in 10% gradients of emotion. See Figure 1 for
2005; Russell, 1980), our goal was to create a series of an example of the series of images.
images spanning both valence and arousal dimensions.
Thus, we aimed to create faces ranging from prototypically Task and Experimental Procedures
happy to prototypically neutral to prototypically angry, each
expressing slight differences in affective expression. The After participants received a general description of the
stimulus set comprised 30 pictures of five male and five fe- experiment and provided informed consent, EEG/EOG sen-
male models each posing angry, happy, and neutral facial sor electrodes were attached and participants were given
expressions taken from Ekman and Friesen’s (1976) Pictures detailed task instructions. Each participant was seated
of Facial Affect. For each model, happy-neutral and angry- approximately 65 cm directly in front of a 17-inch computer
neutral morphs were performed. The happy-neutral morphs monitor with each face stimulus occupying 14.6 degrees of
were created using the happy and the neutral faces to create visual angle vertically and 9.7 degrees of visual angle hori-
a series of nine composite images expressing varying de- zontally. A facial affect categorization task was administered
grees of happiness. The angry-neutral morphs were created on a Pentium class computer, using Presentation software
using the angry and neutral faces to create a series of nine (Neurobehavioral Systems, Albany, CA), to control the pre-
composite images expressing varying degrees of anger. sentation and timing of all stimuli, the determination of
SmartMorph (http://logicnet.dk/SmartMorph/) picture-morp- response accuracy, and the measurement of reaction times.
hing software was used for all morphing procedures. The primary purpose of the facial affect categorization task
was to focus participants’ attention on the facial stimuli. The
The techniques outlined in Steyvers (1999) were used as task required participants to categorize each face as negative,
the basis for morphing the facial stimuli in the current study. positive, or neutral. Prior to the task, all participants were
In preparation for morphing, control points were placed on given detailed instructions and 10 practice trials that
the corresponding regions of both faces (e.g., prototypical included stimuli randomly selected from the larger set of
happy and neutral). The facial regions were defined as the 210 images. During the task, the 210 face images were pre-
eyebrows, eyes, mouth, nose, cheek/chin, forehead, circum- sented in random order two times each across 10 blocks of
ference of face, and hair/neck (see Table 1). The number of 42 trials. To control for repetition effects, each image was
points on each facial region was determined based on meth- presented once during the first half of the study (blocks
ods described in Steyvers and modified for each face as 1–5) and then a second time during the second half of the
needed so as to minimize the amount of blurriness around study (blocks 6–10). A fixation mark (+) was always pre-
the eyes and mouth during morphing. The eyes and mouth sented at the center of the screen during the interstimulus
were given special attention, as a number of studies suggest interval (ISI) to help participants remain focused throughout
that these regions of the face are most important for facial the task. Each face replaced the fixation cross in the center
affect expression and recognition (e.g., Adolphs, 2002; of the computer screen for 1,000 ms against a black back-
Adolphs et al., 2005; Morris, Debones, & Dolan, 2002). ground with an ISI of 1,500 ms. Participants were instructed
Control points were placed on other areas of the face as

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38

30 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

Figure 1. A series of morphs used for the current study. They range from prototypically angry to prototypically neutral
(top) and prototypically happy to prototypically neutral (bottom). Face stimuli were from Ekman and Friesen’s (1976)
Pictures of facial affect.

to categorize the emotion of the face as positive, neutral, or subsequent analyses if: the record contained 25 ms of invari-
negative by pressing the left, down, or right arrow keys with ant (i.e., flat) analog data on any channel or the signal devi-
the right index, middle, and ring fingers, respectively, on a ated by 200 lV above or below the pre-stimulus baseline.
standard keyboard as quickly and as accurately as possible. Single trial EEG data were lowpass filtered at 20 Hz with
The response window was 2,000 ms following face onset. a 51-weight FIR digital filter as per Cook and Miller
Responses made outside this window were discarded. (1992). Finally, the EEG for each trial was time-locked to
At the completion of the facial affect categorization task, stimulus onset and averaged across all levels (21) of
participants rated the facial affect pictures using the SAM. morphed faces ranging from prototypically angry to proto-
typically neutral to prototypically happy. This yielded ERPs
Psychophysiological Recording, Data for each electrode site. After artifact rejection, the mean
Reduction, and Analysis number of averages used to calculate grand averages for
each condition per participant was 19.86 (range = 16–20).
The electroencephalogram (EEG) was recorded from the A 2 (Facial Affect) · 11 (Morph Level) ANOVA revealed
frontal (Fz), fronto-central (FCz), central (Cz), and parietal no significant main or interaction effects, verifying that after
(Pz) recording sites using an ECI electrocap. In addition, artifact rejection, there were no differences in the number of
tin disk electrodes were placed on the left and right mastoids participant averages used to calculate ERP grand averages
(M1 and M2, respectively). During the recording, all activity across conditions, Fs < 1.2, ps > .32. Furthermore, trials
was referenced to Cz. The electrooculogram (EOG) gener- were excluded from all analyses if the subject made an inac-
ated from blinks and vertical eye movements was recorded curate response on the categorization task, resulting in an
using Med-Associates miniature electrodes placed approxi- average of 13.5 trials used to calculate ERP averages for
mately 1 cm above and below the participant’s right eye. each condition.
The right earlobe served as a ground site. All EEG/EOG
electrode impedances were kept below 10 KX and the data Average activity in the 100 ms pre-stimulus window
from all channels were recorded by a Grass Model 78D served as the baseline for all stimulus-locked ERPs. We
polygraph with Grass Model 7P511J preamplifiers (band- were particularly interested in examining positive shifts
pass = 0.1–100 Hz). occurring across time and location. Specifically, we homed
analyses on early frontally-maximal positive shifts (Eimer
All bioelectric signals were digitized on a laboratory & Holmes, 2002), as well as later parietally-maximal posi-
microcomputer using VPM software (Cook, 1999). The tive shifts, referred to collectively as the LPP (Cuthbert
EEG was sampled at 200 Hz. Data collection began et al., 2000; Moser et al., 2008; Olofsson, Nordin, Sequeira,
500 ms prior to the onset of the imperative stimulus and & Polich, 2008; Schupp et al., 2004). We defined three early
continued for 1,500 ms. Off-line, the EEG for each trial time windows occurring 80–300 ms following stimulus on-
was corrected for vertical EOG artifacts using the method set, based on Eimer and Holmes (2002) and the pattern of
developed by Gratton, Coles, and Donchin (1983; Miller, our waveforms. In addition, the LPP was examined in three
Gratton, & Yee, 1988) and then re-referenced to the average parts, as previous studies have suggested that different time
activity of the mastoid electrodes. In line with previous face windows of the LPP indicate different processes related to
studies from our laboratory (e.g., Moser, Huppert, Duval, & attention and memory (Olofsson et al., 2008). These previ-
Simons, 2008), trials were rejected and not counted in ous studies have suggested that earlier LPP windows index
attention and initial memory storage, whereas later windows
index task demands and are sensitive to arousing, and

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing

E. R. Duval et al.: LPP Reflects Differences in Facial Affect 31

motivationally salient, stimuli. Consistent with previous Results
studies, we examined the LPP at three time windows
(Early-LPP: 310–420, Mid-LPP: 450–650, and Late-LPP: Early ERP Windows (80–310 ms Following
650–1,000 ms). Stimulus Onset)

Each portion of the waveform falling within the areas of Stimulus-locked ERPs for prototypical angry, neutral, and
interest was then scored using data from the electrode site happy faces are presented in Figure 2 across the four record-
where it reached maximum amplitude. This was done to ing sites to illustrate the typical look of the waveforms. The
minimize contamination by component overlap (Luck, early ERP windows were maximal at electrode site Fz, so
2005) and to reflect the shift from early frontal positivity this recording site was used for the early ERP analyses.
to late centro-parietal positivity (Eimer & Holmes, 2007). The 2 (Facial Affect: Happy, Angry) · 9 (Morph Level)
ANOVA did not reveal any significant main effects or inter-
Repeated-measures analyses of variance (ANOVAs) actions for any of the early time windows examined
were performed on ERP, SAM, and behavioral (accuracy (Fs < 1, ps > .44).3
and reaction time; RT) measures using SPSS (v. 17.0;
IBM software) general linear model analysis software. LPP Windows
ERP, SAM, and RT were calculated for correct trials only.2
For all ERP scoring windows, as well as SAM ratings, All LPPs were maximal at electrode site Pz, so this record-
Accuracy, and RT, a 2 (Facial Affect: Happy, Angry) · 9 ing site was used for all LPP analyses.
(Morph Level: 100%, 90%, 80%, 70%, 60%, 40%, 30%,
20%, 10% affect) repeated-measures ANOVA was con- Early LPP (310–420 ms)
ducted, where 100% affect represents the prototypical happy
and angry faces. RT data were not normally distributed for The 2 (Facial Affect: Happy, Angry) · 9 (Morph Level)
all conditions, so square root transformations were applied ANOVA revealed a significant main effect of Facial Affect
to all RT data, resulting in normal distributions. Accuracy in the 310–345 window, F(1, 28) = 5.58, p < .03, indicat-
was defined as a response (positive, negative, neutral) that ing that angry faces elicited greater LPP positivity compared
corresponded with the predominant expression depicted in to happy faces overall. No other main effects or interactions
the morphed faces. For example, a correct response on a trial were significant (Fs < 1.4, ps > .21).
with a morph that was 60% happy and 40% neutral was
‘‘positive.’’ Since there is no predominant expression, and Mid-LPP (450–650 ms)
thus no one correct response, on trials with morphs depicting
50% affect (e.g., a 50% happy morph could be categorized Figure 3 shows LPP amplitude across all morph levels
as ‘‘happy’’ or ‘‘neutral’’), they were not considered in these (except the 50% affect condition) at site Pz for the
analyses. Although these conditions were excluded from 450–650 ms window. There was a main effect of Facial
analyses, figures include all levels of facial affect expression Affect, F(1, 28) = 11.80, p < .01, and a main effect of
for illustrative purposes. Morph Level, F(8, 224) = 6.54, p < .001. The main effect
of Facial Affect indicated that angry faces elicited greater
When significant ANOVA effects were detected, we LPP positivity compared to happy faces overall. There
report follow-up polynomial contrasts on the first two trends was a significant linear trend for the effect of Morph Level,
– linear and quadratic – to more thoroughly describe the nat- Flin(1, 28) = 31.58, p < .001, accounting for 58% of the
ure of effects. Greenhouse-Geisser corrections were used for variance. Faces depicting more affect elicited more LPP pos-
all multiple df effects, and alpha was set at .025 for these fol- itivity than faces depicting less amounts of affect. The qua-
low-up contrasts to correct for multiple comparisons dratic trend for Morph Level was not significant. The Facial
(p = .05/2 = .025). Percentage of variance accounted for Affect by Morph Level interaction was not significant,
by each polynomial contrast was calculated as an indicator F < 1.
of effect size. Finally, in order to further examine associa-
tions between measures, averages were computed across
subjects to obtain one value for each measure for each
morph level. Morph level was treated as a continuous vari-
able ranging from prototypically angry to neutral to proto-
typically happy. Pearson’s correlations between measures
were then computed to examine whether ERP, SAM, and
behavioral measures covaried across morph levels. An alpha
level cutoff of .001 was used to correct for multiple
comparisons.

2 All significant main effects and interactions reported using the correct trial dataset were also significant when all (correct and incorrect)
trials were included for analysis.

3 Based on impressions gleaned from Figure 2, a three-level repeated-measures ANOVA was conducted to examine differences in ERP
amplitude for prototypical angry, neutral, and happy expressions at each of the early time windows. There were no significant main effects
(Fs < 3.2, ps > .05), however.

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38

32 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

Figure 2. Stimulus-locked grand average ERP waveforms at each recording site (Fz, FCz, Cz, Pz) for prototypically
angry (red line), neutral (black line), and happy (blue line) faces.

Late-LPP (650–1,000 ms) effect of Facial Affect indicated that happy faces were rated
as more pleasant than angry faces overall. For the effect of
For the Late-LPP window, there was a main effect of Facial Morph Level, the linear trend was significant,
Affect, F(1, 28) = 20.81, p < .001, indicating that angry Flin(1, 28) = 36.67, p < .001, accounting for 83% of the
faces elicited greater LPP positivity compared to happy variance. This linear effect of Morph Level indicated that
faces overall (see Figure 4). The main effect of Morph Level as faces morphed from neutral and approached the emotion
and the Facial Affect by Morph Level interaction were not prototypes, they were rated as happier or angrier as we ex-
significant, ps > .05. pected. The significant Facial Affect · Linear Morph Level
interaction, F(1, 28) = 667.37, p < .001, accounted for 94%
SAM Ratings of the variance, indicating that the amount of affective
expression displayed by the angry faces was more strongly
SAM ratings for the morphed faces ranging from prototyp- related to differences in valence ratings than the amount of
ical happy to prototypical neutral and prototypical angry to affective expression displayed by the happy faces (see Fig-
prototypical neutral are presented in Figure 5, showing va- ure 5). The Facial Affect · Quadratic Morph Level interac-
lence and arousal ratings. The 2 (Facial Affect: Happy, An- tion was also significant, F(1, 28) = 18.28, p < .001, but
gry) · 9 (Morph Level) ANOVA conducted on the valence only accounted for 1% of the variance.
ratings revealed a main effect of Facial Affect,
F(1, 28) = 434.94, p < .001, a main effect of Morph Level, The same analysis conducted on arousal ratings again
F(8, 224) = 11.65, p < .001, and a Facial Affect · Morph revealed a main effect of Morph Level, F(8, 224) = 67.97,
Level interaction, F(8, 224) = 289.56, p < .001. The main p < .001, with a significant linear trend, Flin(1, 28) = 88.02,
p < .001, accounting for 96% of the variance. Faces depict-
ing more affect were rated as more arousing than faces
depicting less affect (see Figure 5). The main effect of Facial

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing

E. R. Duval et al.: LPP Reflects Differences in Facial Affect 33

Figure 3. LPP magnitude for the Mid-LPP (450–650 ms) window at Pz for all gradients in facial expression ranging from
(A) prototypically happy to prototypically neutral and (B) prototypically angry to prototypically neutral. Error bars
indicate standard error of the mean (SEM).

Affect and the Facial Affect by Morph Level interaction angry, or neutral – and was poorer for intermediate levels of
were not significant, Fs < 1. affect. The Facial Affect · Linear Morph Level interaction
was significant, F(1, 28) = 34.65, p < .001, and accounted
Behavioral Data for 28% of the variance. This interaction effect reveals that
participants categorized all faces depicting greater than 50%
Accuracy happy affect with roughly equal accuracy. However, when
categorizing faces along the angry continuum, participants
Results revealed a main effect of Facial Affect, were more accurate as faces depicted increasing levels of
F(1, 28) = 8.28, p < .01, a main effect of Morph Level, angry affect. This suggests that participants had more diffi-
F(8, 224) = 30.87, p < .001, and a Facial Affect · Morph culty distinguishing ambiguous angry expressions from the
Level interaction, F(8, 224) = 31.41, p < .001. The main other expressions (happy, neutral) compared to happy faces,
effect of Facial Affect indicated that, overall, participants which were identified with equal accuracy across levels of
were more accurate when categorizing happy facial expres- affect above 50% (see Figure 6). The Facial Affect · Qua-
sions (M = .80, SD = 0.11) compared to angry facial dratic Morph Level trend was not significant, F < 1.
expressions (M = .76, SD = 0.15). For the main effect of
Morph Level, the linear trend was not significant Reaction Time
(p > .10), but the quadratic trend was, Fquad(1, 28) =
308.34, p < .001, and accounted for 59% of the variance. For RT, there was a main effect of Facial Affect,
This quadratic effect indicated that accuracy was best for F(1, 28) = 479.66, p < .001, a main effect of Morph Level,
faces falling closer to the prototypes – that is, 100% happy, F(8, 224) = 10.96, p < .001, and a Facial Affect · Morph

Figure 4. LPP magnitude for the Late-LPP (650–1,000 ms) window at Pz for all gradients in facial expression ranging
from (A) prototypically happy to prototypically neutral and (B) prototypically angry to prototypically neutral. Error bars
indicate SEM.

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38

34 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

Figure 5. SAM valence and arousal ratings for all gradients in facial expression ranging from (A) prototypically happy to
prototypically neutral and (B) prototypically angry to prototypically neutral. Error bars indicate SEM.

Level interaction, F(8, 224) = 109.05, p < .001. The main when categorizing faces along the happy continuum, partic-
effect of Facial Affect indicated that, overall, participants ipants were faster for faces depicting high degrees of happi-
were faster at categorizing happy facial expressions ness. When categorizing faces along the angry continuum,
(M = 35.13; SD = 0.73) compared to angry facial expres- however, participants were faster for faces depicting less
sions (M = 36.34; SD = 0.67). For the main effect of Morph anger and more neutral expressions. In addition, individuals
Level, the linear trend was significant, Flin(1, 28) = 12.92, were faster at categorizing happy faces compared to angry
p = .001 and accounted for 45% of the variance. Partici- faces, but this effect was most pronounced for the faces
pants were fastest to categorize the prototypical faces, and depicting greater levels of affect – that is, faces containing
were slower to categorize faces as they became more 60% or more of the happy or angry expression (see
ambiguous. The quadratic trend was also significant, Figure 6).4
Fquad(1, 28) = 37.09, p < .001, but only accounted for
22% of the variance. The Facial Affect · Linear Morph Correlations
Level interaction was significant, F(1, 28) = 531.57,
p < .001, and accounted for 81% of the variance. The Facial The Mid-LPP was positively correlated with SAM arousal rat-
Affect · Quadratic Morph Level trend was also significant, ings (r = .75, p < .001), suggesting an arousal effect in the
F(1, 28) = 5.97, p < .025, but only accounted for less than Mid-LPP. The Mid-LPP followed the same pattern as SAM
1% of the variance. This pattern of effects suggests that

4 In line with the RT data, more timeouts (RTs > 2,000 ms) occurred on trials when more angry expressions were depicted. Specifically,
more timeouts occurred as expressions moved from prototypically angry to neutral, and timeouts decreased as expressions moved from
prototypically neutral to happy. This suggests that participants were so slow to respond to the most angry expressions that they were less
able to make a response within the allotted time window of 2,000 ms.

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing

E. R. Duval et al.: LPP Reflects Differences in Facial Affect 35

Figure 6. Proportion of accurate responses and RT for all gradients in facial expression ranging from (A) prototypically
happy to prototypically neutral and (B) prototypically angry to prototypically neutral. Error bars indicate SEM.

arousal ratings, with greater LPP activation and higher arousal mediated the relationship between the Late-LPP and behav-
ratings as faces increased in their level of affective expression ioral performance.
from neutral toward both ends of the continuum (happy and
angry). The Late-LPP was negatively correlated with SAM Discussion
valence ratings (r = À.83, p < .001) and positively correlated
with RT (r = .74, p < .001). RT and SAM valence ratings The primary aim of the current study was to investigate the
were also negatively correlated with each other (r = À.91, sensitivity of early and late positive shifts of the ERP to
p < .001). These correlations suggest a valence effect for both degrees of facial affect expression. Results revealed that
the Late-LP and RT. As faces increased across the continuum the LPP and behavioral measures were modulated by small
from happy to neutral to angry: differences in the level of facial affect expression. As the
(1) the Late-LPP increased, amount of affective expression of both happy and angry
(2) faces were rated as less positive, and faces increased, so did their motivational relevance, as
(3) RT increased. indexed by enhanced LPP magnitude between 450 and
650 ms post-stimulus onset, more extreme SAM ratings
A partial correlation was conducted to follow up the three- consistent with level of affect expressed, and changes in
way correlation between Late-LPP, SAM valence, and RT. RT and Accuracy. Correlations between measures further
When controlling for SAM valence ratings, the correlation suggested that ERP and behavioral data covaried across
between Late-LPP and RT was no longer significant
(r = .03, p = .90), suggesting that perceived pleasantness

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38

36 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

levels of facial affect expression. Specifically, two processes more thoroughly than other types of stimuli (see Ohman
were observed: the Mid-LPP showed an arousal effect, & Mineka, 2001 for a review). However, these findings
whereas the Late-LPP showed a valence effect. In addition, are inconsistent with previous findings that negative faces
angry faces elicited larger positive amplitudes across all are detected more efficiently during visual search tasks com-
three of the LPP time windows (i.e., 310–345, 450–650, pared to positive faces (e.g., Fox et al., 2000; Horstmann &
and 650–1,000 ms). Bauland, 2006). Although seemingly inconsistent with the
visual search findings (Fox et al., 2000; Horstmann &
The current finding that the LPP varied as a function of Bauland, 2006), the tendency to allocate more attentional re-
the level of affective expression is consistent with previous sources to angry faces may be contributing to slower and
findings that the LPP is larger to emotional compared to less accurate task performance on the timed categorization
neutral faces (Eimer & Holmes, 2002), and is tightly cou- task, as more extensive cognitive evaluation is required be-
pled to increases in the level of activation (i.e., arousal) elic- fore a response can be made (Baumeister et al., 2001). This
ited by emotional scenes (Olofsson et al., 2008). Our SAM notion is consistent with a number of studies suggesting that
findings are consistent with the LPP results and previous re- negative stimuli engage more attentional resources than
ports that healthy participants are sensitive to the affective other types of stimuli, subsequently interfering with perfor-
expression portrayed by prototypical (e.g., Stenberg, mance on cognitive and behavioral tasks (e.g., Stroop Task;
Wiking, & Dahl, 1998) and ambiguous (Thomas et al., McKenna & Sharma, 1995; Go/No-Go task; De Houwer &
2007) emotional expressions. Given that previous studies fo- Tibboel, 2010; a facial feature counting task; Eastwood,
cused mainly on how prototypical faces modulate the LPP Smilek, & Merikle, 2003; and categorization/recognition
(e.g., Eimer & Holmes, 2002), the current LPP and SAM tasks; Leppanen & Hietanen, 2003; Leppanen, Tenhunen,
arousal results extend the literature to suggest that the brain & Hietanen, 2003; Moser et al., 2008). Although some have
is tuned to make fine discriminations between faces depict- suggested that this behavioral advantage for happy faces is
ing slight differences in the level of facial affect among more the result of distinctive physical features (e.g., a smiling
ambiguous, ecologically valid faces, as early as 450 ms. mouth; Adolphs, 2002), others suggest that it is the result
of emotional and cognitive processing differences per se
In addition to the main finding that ERP and behavioral (e.g., Leppanen & Hietanen, 2003).
measures were sensitive to slight changes in facial affect,
our results also support the negativity bias hypothesis. Con- Our findings and previous literature suggest a two-
sistent with previous research (Balconi & Lucchiari, 2005; process model for affective face processing. Initially, the
Balconi & Pizzoli, 2003; Dennis & Chen, 2007; Eger, Mid-LPP was correlated with SAM arousal ratings, demon-
Jedynak, Iwaki, & Skrandies, 2003; Eimer et al., 2003; strating an arousal effect, such that more attentional
Leppanen, Moulson, Vogel-Farley, & Nelson, 2007; Liddell, resources were directed toward the most activating or arous-
Williams, Rathjen, Shevrin, & Gordon, 2004; Schupp et al., ing stimuli, as they were the most motivationally salient
2004), the LPP was enhanced in all time windows (310– (Pratto & John, 1991). Then, the Late-LPP was correlated
1,000 ms) during the processing of angry compared to hap- with SAM valence ratings, demonstrating a valence effect,
py faces across all levels of expression. Angrier faces were such that negative faces, but not positive faces, continued
also associated with impaired behavioral performance, as to engage attentional resources (Fox et al., 2000) and were
indicated by slower RTs and lower accuracy than happier processed more thoroughly (Baumeister et al., 2001). The
faces. Although happy faces tended to be perceived as correlation between Late-LPP and RT was mediated by
equally happy across levels of affect intensity, participants SAM valence ratings, suggesting that perceived pleasantness
provided more graded responses to angry faces, relative to was related to behavioral performance, such that as more
the amount of affect expressed. attentional resources were allocated to the more negatively
perceived faces, the system slowed down and behavioral
Together with the behavioral findings, one might argue performance was impaired.
that angry faces may have elicited larger LPP amplitude
because they were more difficult to categorize than happy Our results did not replicate previous reports that affec-
faces. However, if difficulty was driving this effect, we tive faces result in a frontal positive shift in early ERP win-
would expect the more ambiguous angry faces to have the dows. The lack of enhanced positivity to affective faces
largest LPP, as they would be the most challenging to cate- across earlier components is consistent with Schupp et al.
gorize. Given that the largest LPP values were observed for (2004), who however, demonstrated no reliable modulation
angry faces displaying the greatest amounts of affective of any ERP components in response to affective faces prior
expression, we can thus disassociate difficulty from emo- to 200 ms following face onset. Previous research investi-
tion, and conclude that these effects were a function of the gating modulations of early ERP positivity by affective
motivational salience/emotional qualities of the stimuli. That expressions embedded the face stimuli among other objects
the happy and angry faces in the current study did not differ (e.g., houses) and showed generic positivity enhancements
on overall SAM-reported arousal levels rules out the possi- across several different expressions (Eimer & Holmes,
bility of an arousal level confound. We have strong support, 2007). Therefore, the lack of early ERP modulation in the
then, that angry faces were both harder to classify than current study is not surprising given that all of our stimuli
happy faces, and attracted more attentional resources during were faces – as was also the case in Schupp et al. Together
the LPP time windows than did happy faces. with the LPP results, these findings suggest that components
reflecting later, more complex, processes of the visual
The current findings are consistent with reports that neg- system play the biggest role in affective discrimination.
ative stimuli attract more attentional resources (Baumeister,
Bratslavsky, Finkenauer, & Vohs, 2001) and are processed

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing

E. R. Duval et al.: LPP Reflects Differences in Facial Affect 37

This is especially evident given that the LPP, originating in Adolphs, R. (2002). Recognizing emotion from facial expres-
parietal, inferotemporal, and occipital regions (Sabatinelli, sions. Psychological and neurological mechanisms. Behav-
Lang, Keil, & Bradley, 2007; Schupp et al., 2000), showed ioral and Cognitive Neuroscience Reviews, 1, 21–62.
the greatest sensitivity to changes in facial affect expression.
Given the interplay between frontal, visual, and subcortical Adolphs, R., Gosselin, F., Buchanan, T. W., Tranel, D., Schyns,
regions during affective face processing (Eimer & Holmes,
2007), future studies using dense electrode montages as well P., & Damasio, A. R. (2005). A mechanism for impaired fear
as functional magnetic resonance imaging (fMRI) are war- recognition after amygdala damage. Nature, 433, 68–72.
ranted in order to further elucidate these interactive and dy-
namic mechanisms. Balconi, M., & Lucchiari, C. (2005). Event-related potentials
related to normal and morphed emotional faces. The Journal
To extend the current findings, future research could also of Psychology, 139, 176–192.
examine the effects of individual differences, particularly
those characterized by an oversensitivity to social threat Balconi, M., & Pizzoli, U. (2003). Face-selective processing and
(i.e., social anxiety and depression), on ERPs and behavioral
performance during morphed face processing. Prototypically the effect of pleasant and unpleasant emotional expressions
negative facial expressions are typically used in studies on ERP correlates. International Journal of Psychophysiol-
investigating face processing in social anxiety (e.g., Moser ogy, 49, 67–74.
et al., 2008). However, given that people with social anxiety
tend to interpret ambiguous faces as negative (Franklin, Baumeister, R. F., Bratslavsky, E., Finkenauer, C., & Vohs, K. D.
Huppert, Langner, Leiberg, & Foa, 2005; Hirsch & Clark, (2001). Bad is stronger than good. Review of General
2004), future studies could extend this research to examine Psychology, 5, 323–370.
group differences in the processing of a variety of levels of
facial affect. The preliminary study by Cavanagh and Bradley, M. M. (2009). Natural selective attention: Orienting and
Geisler (2006) in depressed undergraduates provides initial emotion. Psychophysiology, 46, 1–11.
support for examining ERP responses to morphed faces in
groups showing sensitivity to social signals. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The
self-assessment manikin and the semantic differential. Jour-
Overall, the findings described in the present study lead nal of Behavioral Therapy and Experimental Psychiatry, 25,
to two main conclusions regarding how healthy participants
respond to facial affect expression. First, healthy participants 49–59.
discriminate between slight differences in the level of facial
affect. An arousal effect was demonstrated in the Mid-LPP, Campanella, S., Quinet, P., Bruyer, R., Crommelinck, M., &
such that faces depicting greater levels of affective expres-
sion elicited enhanced Mid-LPP magnitude and greater Guerit, J. M. (2002). Categorical perception of happiness and
arousal ratings, while LPP magnitude and arousal ratings fear facial expressions: An ERP study. Journal of Cognitive
decreased as faces became more neutral. Second, a valence Neuroscience, 14, 210–227.
effect was observed in the Late-LPP, suggesting that angry
faces continue to engage more processing resources than Cavanagh, J., & Geisler, M. W. (2006). Mood effects on the ERP
happy faces. This continued attention allocation toward
angry faces was related to poorer response accuracy and processing of emotional intensity in faces: A P3 investigation
slower RT. Together, the current findings highlight the with depressed students. International Journal of Psycho-
brain’s sensitivity to the salience of social signals and sug- physiology, 60, 27–33.
gest a two-process model of affective face processing. Cook, E. W. III (1999). VPM reference manual. Birmingham,
AL: Author.
Acknowledgments
Cook, E. W. III, & Miller, G. A. (1992). Digital filtering:
The authors would like to thank Ashley Malooly and Natalie Background and tutorial for psychophysiologists. Psycho-
Shroyer for their help with data collection. The authors physiology, 29, 350–367.
would also like to thank Eva Gilboa-Schechtman for provid-
ing stimuli. This research was supported by National Insti- Cuthbert, B. N., Schupp, H. T., Bradley, M. M., Birbaumer, N.,
tute of Mental Health (NIMH) predoctoral fellowship
MH077388 (JSM). & Lang, P. J. (2000). Brain potentials in affective picture

References processing: Covariation with autonomic arousal and affective
report. Biological Psychology, 52, 95–111.
Achaibou, A., Pourtois, G., Schwartz, S., & Vuilleumier, P.
(2008). Simultaneous recording of EEG and facial muscle De Houwer, J., & Tibboel, H. (2010). Stop what you are doing!
reactions during spontaneous emotional mimicry. Neuro-
psychologia, 46, 1104–1113. Emotional pictures interfere with the task not to respond.
Psychonomic Bulletin and Review, 17, 699–703.

Dennis, T. A., & Chen, C. C. (2007). Emotional face

processing and attention performance in three domains:

Neurophysiological mechanisms and moderating effects of
trait anxiety. International Journal of Psychophysiology,
65, 10–19.
Donchin, E. (1981). surprise! . . . Surprise? Psychophysiology,
18, 493–513.
Donchin, E., & Coles, M. G. H. (1988). Is the P300 component a
manifestation of contextual updating? Behavioral and Brain
Sciences, 11, 357–427.

Eastwood, J. D., Smilek, D., & Merikle, P. (2003). Negative

facial expression captures attention and disrupts perfor-
mance. Attention, Perception, and Psychophysics, 65, 852–

858.

Eger, E., Jedynak, A., Iwaki, T., & Skrandies, W. (2003). Rapid

extraction of emotional expression: Evidence from evoked

potential fields during brief presentation of face stimuli.
Neuropsychologia, 41, 808–817.

Eimer, M., & Holmes, A. (2002). An ERP study on the time
course of emotional face processing. Neuroreport, 13, 4959–

4965.

Eimer, M., & Holmes, A. (2007). Event-related brain potential
correlates of emotional face processing. Neuropsychologia,
45, 15–31.

Eimer, M., Holmes, A., & McGlone, F. P. (2003). The role of
spatial attention in the processing of facial expression:

An ERP study of rapid brain responses to six basic

Hogrefe Publishing Journal of Psychophysiology 2013; Vol. 27(1):27–38

38 E. R. Duval et al.: LPP Reflects Differences in Facial Affect

emotions. Cognitive, Affective, & Behavioral Neuroscience, Ohman, A., & Mineka, S. (2001). Fears, phobias, and prepared-
3, 97–110. ness: Toward an evolved module of fear and fear learning.
Ekman, P., & Friesen, W. V. (1976). Pictures of facial affect. Palo Psychological Review, 108, 483–522.

Alto, CA: Consulting Psychologists Press. Olofsson, J. K., Nordin, S., Sequeira, H., & Polich, J. (2008).
Affective picture processing: An integrative review. Biolog-
Fox, E., Lester, V., Russo, R., Bowles, R. J., Pichler, A., & ical Psychology, 77, 247–265.

Dutton, K. (2000). Facial expressions of emotions: Are angry Peeters, G., & Czapinski, J. (1990). Positive-negative asymmetry
faces detected more efficiently? Cognition and Emotion, 14, in evaluations: The distinction between affective and informa-
tional negativity effects. In W. Stroebe & M. Hewstone (Eds.),
61–92. European review of social psychology (Vol. 1, pp. 33–60).
Chichester, UK: Wiley.
Franklin, M. E., Huppert, J. D., Langner, R., Leiberg, S., & Foa,
Posner, J., Russell, J. A., & Peterson, B. S. (2005). The
E. B. (2005). Interpretation bias: A comparison of treated circumflex model of affect: An integrative approach to
affective neuroscience, cognitive development, and psycho-
social phobics, untreated social phobics, and controls. pathology. Development and Psychopathology, 17, 715–
Cognitive Therapy and Research, 29, 289–300. 734.
Gratton, G., Coles, M. G. H., & Donchin, E. (1983). A new
method for off-line removal of ocular artifact. Electroen- Pratto, F., & John, O. P. (1991). Automatic vigilance: The
cephalography and Clinical Neurophysiology, 55, 468–484. attention-grabbing power of negative social information.
Journal of Personality and Social Psychology, 61, 380–
Hirsch, C. R., & Clark, D. M. (2004). Information-processing 391.
bias in social phobia. Clinical Psychology Review, 24, 799–
Russell, J. A. (1980). A circumflex model of affect. Journal of
825. Personality and Social Psychology, 39, 1161–1178.

Horstmann, G., & Bauland, A. (2006). Search asymmetries with Sabatinelli, D., Lang, P. J., Keil, A., & Bradley, M. M. (2007).
real faces: Testing the anger superiority effect. Emotion, 6, Emotional perception: Correlation of functional MRI and
event related potentials. Cerebral Cortex, 17, 1085–1091.
193–207.
Schupp, H. T., Cuthbert, B. N., Bradley, M. M., Cacioppo, J. T.,
Ito, T. A., Larsen, J. T., Smith, N. K., & Cacioppo, J. T. (1998). Ito, T., & Lang, P. J. (2000). Affective picture processing:
The late positive potential is modulated by motivational
Negative information weighs more heavily on the brain: The relevance. Psychophysiology, 37, 257–261.
negativity bias in evaluative categorizations. Journal of
Personality and Social Psychology, 75, 887–900. Schupp, H. T., Ohman, A., Junghofer, M., Weike, A. I.,
Stockburger, J., & Hamm, A. O. (2004). The facilitated
Kanouse, D. E., & Hansen, L. R., Jr. (1971). Negativity in processing of threatening faces: An ERP analysis. Emotion,
4, 189–200.
evaluations. In E. E. Jones, D. E. Kanouse, H. H. Kelley, R.
E. Nisbett, S. Valin, & B. Weiber (Eds.), Attribution: Skowronski, J. J., & Carlston, D. E. (1989). Negativity and
Perceiving the causes of behavior (pp. 47–62). Morristown, extremity biases in impression formation: A review of
explanations. Psychological Bulletin, 105, 131–142.
NJ: General Learning Press.
Sprengelmeyer, R., & Jentzsch, I. (2006). Event related potentials
Leppanen, J. M., & Hietanen, J. K. (2003). Affect and face and the perception of intensity in facial expressions. Neuro-
perception: Odors modulate the recognition advantage of psychologia, 44, 2899–2906.
happy faces. Emotion, 3, 315–326.
Stenberg, G., Wiking, S., & Dahl, M. (1998). Judging words at
Leppanen, J. M., Moulson, M. C., Vogel-Farley, V. K., & face value: Interference in a word processing task reveals
automatic processing of affective facial expressions. Cogni-
Nelson, C. A. (2007). An ERP study of emotional face tion and Emotion, 12, 755–782.
processing in the adult and infant brain. Child Develop-
ment, 78, 232–245. Steyvers, M. (1999). Morphing techniques for manipulating face
images. Behavior, Research Methods, Instruments and Com-
Leppanen, J. M., Tenhunen, M., & Hietanen, J. K. (2003). Faster puters, 31, 359–369.

choice-reaction times to positive than to negative facial Thomas, L. A., De Bellis, M. D., Graham, R., & LaBar, K. S.
expressions. Journal of Psychophysiology, 17, 113–123. (2007). Development of emotional facial recognition in late
childhood and adolescence. Developmental Science, 10, 547–
Liddell, B. J., Williams, L. M., Rathjen, J., Shevrin, H., & 558.

Gordon, E. (2004). A temporal dissociation of subliminal Tipples, J., Atkinson, A. P., & Young, A. W. (2002). The
eyebrow frown: A salient social signal. Emotion, 2, 288–296.
versus supraliminal fear perception: An event-related poten-
tial study. Journal of Cognitive Neuroscience, 16, 479–486. Accepted for publication: August 29, 2012
Luck, S. J. (2005). An introduction to the event-related potential
technique. Cambridge, MA: MIT Press. Elizabeth R. Duval

Miller, G. A., Gratton, G., & Yee, C. M. (1988). Generalized Department of Psychiatry
University of Michigan
implementation of an eye movement correction procedure. 4250 Plymouth Rd.
Psychophysiology, 25, 241–243. Ann Arbor, MI 48109
McKenna, F. P., & Sharma, D. (1995). Intrusive cognitions: An USA
investigation of the emotional Stroop Task. Journal of Tel. +1 816 206-6555
Experimental Psychology: Learning, Memory, & Cognition, E-mail [email protected]
21, 1595–1607.

Morris, J. S., Debones, M., & Dolan, R. J. (2002). Human
amygdala responses to fearful eyes. NeuroImage, 17, 214–

222.

Moser, J. S., Huppert, J. D., Duval, E., & Simons, R. F. (2008).

Face processing biases in social anxiety: An electrophysio-
logical study. Biological Psychology, 78, 93–103.

Nieuwenhuis, S., Aston-Jones, G., & Cohen, J. D. (2005).

Decision making, the P3, and the locus coeruleus norepi-
nephrine system. Psychological Bulletin, 131, 510–532.

Ohman, A., Lundqvist, D., & Esteves, F. (2001). The face in

the crowd revisited: A threat advantage with schematic
stimuli. Journal of Personality and Social Psychology, 80,

381–396.

Journal of Psychophysiology 2013; Vol. 27(1):27–38 Hogrefe Publishing